Re: Bug#218081: [peterc@gelato.unsw.edu.au: Bug#218081: libc6 2.3.2-9 won't install]
On Thu, Oct 30, 2003 at 10:38:57AM +1100, Peter Chubb wrote:
> >>>>> "Daniel" == Daniel Jacobowitz <dan@debian.org> writes:
>
> Daniel> Meant to copy yo on this, Peter. Adam suggests checking the
> Daniel> output of lsattr. If you can repeat this, strace output of
> Daniel> dpkg would be nice too.
>
> The root filesystem is reiserfs.
>
> lsattr /lib/ld-2.3.2.so
> s-------c-------- ld-2.3.2.so
Do those mean the same things they do on ext2? If so, they're "zero
when deleted" and "transparent compression". Bizarre.
This sounds like a kernel bug, not a dpkg or libc bug.
>
> And this is the relevant part of strace output:
> ...
>
> [pid 1280] utime("/lib/ld-2.3.2.so.dpkg-new", [2003/10/30-10:28:13, 2003/10/27-13:34:17]) = 0
> [pid 1280] link("/lib/ld-2.3.2.so", "/lib/ld-2.3.2.so.dpkg-tmp") = 0
> [pid 1280] rename("/lib/ld-2.3.2.so.dpkg-new", "/lib/ld-2.3.2.so") = -1 EBUSY (
> Device or resource busy)
Rename is not documented as returning EBUSY normally. Investigate why
this happened.
> [pid 1280] write(2, "dpkg: error processing libc6_2.3"..., 138dpkg: error proce
> ssing libc6_2.3.2-9_i386.deb (--install):
> unable to install new version of `./lib/ld-2.3.2.so': Device or resource busy
> ) = 138
> [pid 1280] lstat64("//lib/ld-2.3.2.so.dpkg-tmp", {st_mode=S_IFREG|0755, st_size=92174, ...}) = 0
> [pid 1280] rename("//lib/ld-2.3.2.so.dpkg-tmp", "//lib/ld-2.3.2.so") = 0
> [pid 1280] rmdir("//lib/ld-2.3.2.so.dpkg-new") = -1 ENOTDIR (Not a directory)
> [pid 1280] unlink("//lib/ld-2.3.2.so.dpkg-new") = 0
>
>
> I note that /lib/ld-2.3.2.so.dpkg-tmp is left behind after the process
> completes, too.
>
--
Daniel Jacobowitz
MontaVista Software Debian GNU/Linux Developer
Reply to: